AI in small business marketing: what to hand to the tool and what to keep

In a small company, marketing is usually done by the owner or by one person “on the side”. AI tools look ideal: a post in a minute, a service description in two, a newsletter in five. After a few weeks, though, it turns out that all the texts sound alike - and alike to the competition’s texts, because they use the same tools.

The problem is not the tool. It is what goes into it.

Why AI texts all sound the same

A tool that knows nothing about your company writes an average text - correct, smooth and fitting any company in the industry. Words like “comprehensive”, “professional” and “tailored approach” appear because they fit everywhere. A client reads such a text and does not know why they should choose you.

A differentiator will not come out of the tool. You have to bring it.

Before you start: one page about your brand

Prepare a short document to attach to every request:

• who you help - specifically, not “businesses and individuals”,
• with what - which problems you solve,
• how you differ - based on facts, not adjectives,
• how you speak - formal or informal, brief or detailed, with humour or without,
• what you do not promise - so the tool does not add “guarantees” and “the best price”,
• two or three texts you consider good - as a model for tone.

If the first three points are hard to answer, that is a sign that the communication itself needs putting in order before AI can help. That is what I do as part of brand and image work.

Where AI genuinely helps

• Variants of one text for different channels. A good service description becomes a post, a newsletter paragraph and a short version for a business listing.
• Summaries and reworking of longer content. An article becomes a few posts; a call recording becomes a list of questions worth answering on the website.
• First drafts of replies to common client questions, based on your real answers.
• Topic ideas based on the questions clients actually ask.
• Proofreading and simplifying the language of texts you have already written.
• Descriptions following a template - for example a series of product descriptions in the same layout.

What they have in common: the tool works on material that already exists and carries your knowledge.

What not to hand over to the tool

• Your brand’s differentiator. The tool can help you phrase it, but it will not invent how the company really differs.
• Promises and numbers. Every number, deadline and promise is checked by a person. I describe a simple approach in the article on checking AI answers.
• Client reviews and recommendations. These are never generated. A made-up review misleads clients, however good it sounds.
• Images of people. A generated “team photo” or “happy client” pretends to be something that does not exist.
• Replies to difficult public comments. A draft, yes, but a person who understands the context sends it.

AI-generated images

Generated illustrations can be useful as a background, an icon or a post graphic. Two rules: check the tool’s licence terms for commercial use, and do not present generated images as photos of real people, places or projects.

A simple process that keeps quality up

1. Brief - the purpose of the text, the audience, the channel, the one page about your brand.
2. A draft from the tool.
3. Editing by a person - does it sound like you and does it say something specific.
4. Fact checking - numbers, promises, names.
5. Publishing.
6. Saving good texts as templates - every month the tool gets better input.

What to measure

Instead of asking whether AI “works in marketing”, measure what can be measured: how long it takes to prepare a typical piece of content, how many corrections a draft needs and whether the number of client enquiries changes. The last measure depends on many factors at once, so treat it as an observation, not proof.